Microsoft's report on AI adoption across the US.

The San Francisco Board of Supervisors postponed voting on a 362-page ordinance that would make over a hundred changes to the city code, after several supervisors and the board clerk raised concerns that the proposal was too broad and unclear. The ordinance, a product of AI research with Stanford University, aims to remove redundant and obsolete reporting requirements.

Agentic OS is a proactive AI assistant that seamlessly automates tasks, scheduling, and files.

The author built a local text-to-speech app StreamTTS using NVIDIA Jetson Orin Nano Super and Kokoro-82M, replacing traditional request-response with durable streams (S2) for shareable, live-updating audio generation, addressing slow inference, fair scheduling, and deduplication.

Sierra's Head of Agent Engineering Natalie Meurer discusses the evolution of forward deployed engineering, arguing that the role is defined by customer accountability rather than specific skills. She notes convergence with product engineering and the rise of agent engineers.

This article describes a new method using Automatic Identification System (AIS) heading data to track grain smuggling from occupied Ukraine to Libya. In February 2026, the bulk carrier Grumant loaded grain at the occupied port of Feodosia in Crimea and after a two-month journey arrived in Benghazi, Libya. Due to GNSS interference in the Black Sea, AIS position data is unreliable, but heading data from the ship's compass is unaffected and can be used to confirm location. By analyzing Grumant's heading data alongside satellite imagery, researchers confirmed its presence at Feodosia. This technique provides a new tool for tracking illicit shipping.

The U.S. Commerce Department has lifted export controls on Anthropic's Fable 5 and Mythos 5 AI models, allowing Fable to return to users. Anthropic agreed to proactively detect and address security risks and work with the U.S. government on protocols. The controls were imposed over two weeks ago due to a 'jailbreak' concern, causing industry unease.

After weeks of negotiations with the Trump administration, Anthropic is finally able to bring Claude Fable 5 back online. Access restoration is set to begin tomorrow.

Multi-token Residual Prediction (MRP) is a lightweight module for diffusion language models that predicts the residual between adjacent denoising steps rather than the full distribution. This enables both lossless speedup in static regimes (up to 1.56×) and quality recovery in dynamic regimes (up to +16 accuracy points) without trade-offs.

Hugging Face and Cerebras have collaborated to create a real-time voice AI system powered by Gemma 4, achieving dramatically lower latency through an open modular architecture. The pipeline integrates Nvidia's speech recognition, Cerebras's fast inference, and Alibaba's text-to-speech, and is already deployed in over 9,000 Reachy Mini robots.

Together AI raised $800M in Series C funding to accelerate the shift to open-source AI. The company argues that closed models' economics don't scale, and open models combined with full-stack optimization can achieve 6-20x cost reductions. Together AI has launched innovations like FlashAttention-4 and Together Megakernel, becoming one of the world's largest AI token producers.

Anthropic announces that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5, with access restoration starting tomorrow.

RunInfra automatically optimizes open-source AI models for production by selecting the best inference engine, GPU, and configuration through benchmarking and tuning, delivering a deployable stack with significant latency, throughput, and cost improvements.

After two packed AIEWF workshops, Ahmad Osman makes the case that local AI is catching up fast — from laptops and phones to enterprise-grade infrastructure. Open-source models are narrowing the gap with proprietary frontier models, and the local AI ecosystem is expanding beyond models to include complete tooling and infrastructure layers.

Claude Sonnet 5 scores 53 on the Artificial Analysis Intelligence Index, matching GPT-5.5 with high reasoning but behind Opus 4.7 and 4.8. It costs $2.29 per task, ~15% more than Opus 4.8, driven by increased token usage. Sonnet 5 matches or outperforms Opus 4.8 on agentic knowledge work tasks but lags on reasoning-heavy benchmarks. It retains a 1M token context window and adds an 'xhigh' effort setting.

This article examines AI's impact on higher education, comparing it to previous technological shocks such as the Internet, big data, and MOOCs. The author argues that while these earlier shocks had limited effects on core educational practices, AI's impact is more profound, broad, systemic, and rapid, potentially transforming the very nature of learning and teaching.

Anthropic has announced a new AI drug discovery program, aiming to accelerate pharmaceutical research with AI.

Despite media hype that AI will replace mathematicians, AI cannot verify the correctness of mathematical proofs; humans remain the ultimate arbiters of correctness.

AI capabilities are accelerating faster than exponential, with frontier and open-weight models both improving rapidly. Usage is shifting from chatbots to autonomous agents, and domain expertise beats profession. Exponential growth causes instability that won't settle soon.

The author argues that most numbers we encounter in news, reports, and daily work are wrong or misleading. Citing examples from the New York Times misstating global COVID cases by a factor of 1,000, confusing wheat production with exports, and workplace errors involving cost overruns and unit mix-ups, the essay builds a case for default skepticism. The underlying message: numbers are powerful but dangerous shortcuts; real understanding requires effort and multiangle verification.

Google releases Nano Banana 2 Lite (aka Gemini 3.1 Flash Lite Image), billed as the fastest and cheapest Gemini image model. The author tested it with a 'Where's Waldo' prompt and got good results despite spelling errors.

Despite AI's potential, public sentiment is overwhelmingly negative. This article explores historical tech skepticism, five reasons for AI's unique hatred, and strategies to improve AI's public image.

The Compute Tightness Index dropped to 43.8/100 as of 2026-06-29, entering Loose territory with a -13.7 pt monthly decline. Spot discounts on H100, B200, and H200 reached 55%, 47%, and 31%, signaling abundant idle AI infrastructure and supply exceeding near-term demand.

Anthropic has launched Claude Sonnet 5, its most agentic mid-tier model, outperforming Sonnet 4.6 across all benchmarks and narrowing the gap to Opus 4.8. It introduces effort levels to control reasoning costs, offering great value at low/medium effort but potentially exceeding Opus 4.8 cost at extra-high effort. It is now the default model for Free and Pro plans and accessible via API.

Anthropic is not the first AI vendor to focus on science and is proceeding cautiously due to the challenges of using AI in the scientific field.

FluidVoice is an open-source macOS dictation app that provides local AI-enhanced speech-to-text. It supports multiple speech models, command mode, write mode, and more, all running on-device for privacy. The latest 1.6.0 release introduces ultra-low latency Parakeet engine and Fluid Intelligence local AI enhancement.

Anthropic releases Claude Sonnet 5, with performance close to Opus 4.8 at lower prices. The model introduces a new tokenizer that increases token count by approximately 30% for the same input, effectively raising costs. API changes include removal of sampling parameters and adaptive thinking enabled by default.

Anthropic has released Claude Sonnet 5, the mid-range model free for all users. It features improved task follow-through, self-verification, and agentic tool use, making it faster, cheaper, and ideal for everyday tasks.

The Godot Foundation announces new contribution guidelines forbidding AI-authored code, AI-agent pull requests, and AI-generated text in human communication, citing the draining and demoralizing effect of low-quality AI slop on maintainers.

Google announces major updates to Gemini Spark, including a macOS desktop app, integrations with Google Tasks, Keep, Canva, Dropbox, Instacart, OpenTable, and Zillow Rentals, and real-time topic tracking. The macOS app automates file and app tasks across the desktop with remote execution coming soon. Spark can now intelligently monitor sports, finance, news, and more.